Expression patterns of cyclin D1 and relatedproteins regulating G1-S phase transition in uvealmelanoma and retinoblastoma

S E Coupland, N Bechrakis, A Schüler, I Anagnostopoulos, M Hummel, N Bornfeld,H Stein

AbstractBackground/aims—A checkpoint mech-anism in late G1, whose regulation vialoss of retinoblastoma protein (pRB) orp16, or overexpression of cyclin D1 orcyclin dependent kinase 4 (CDK4), hasbeen proposed to constitute a commonpathway to malignancy. The aims of thisstudy were (a) to compare markers of cellcycle G1-S phase transition in an in-traocular tumour with known pRB defi-ciency (retinoblastoma) and compare itwith one with an apparently functionalpRB (uveal melanoma); (b) to determineif one of these markers may have a role inthe pathogenesis of uveal melanoma; and(c) to determine if there is a diVerence incell cycle marker expression followingtreatment of uveal melanoma and retino-blastoma.Methods—90 eyes were enucleated from89 patients for retinoblastoma (n=24) orfor choroidal or ciliary body melanoma(n=66). Conventional paraYn sectionswere assessed for cell type and degree ofdiVerentiation. Additional slides wereinvestigated applying standard immuno-histochemical methods with antibodiesspecific for cyclin D1 protein, pRB, p53,p21, p16, BCL-2, and MIB-1.Results—Cyclin D1 protein and pRB werenegative in retinoblastoma using the ap-plied antibodies. In contrast, cyclin D1protein expression was observed in 65% ofuveal melanomas; a positive correlationbetween cyclin D1 cell positivity andtumour cell type, location, growth frac-tion, as well as with pRB positivity wasobserved. p53, p21, and p16 could bedemonstrated in both tumours. An inverserelation between p53 and p21 expressionwas demonstrated in most choroidalmelanomas and in some retinoblastomas.Apart from a decrease in the growth frac-tions of the tumours as determined byMIB-1, a significant diVerence in theexpression of G1-S phase transition mark-ers in vital areas of uveal melanoma andretinoblastoma following treatment withradiotherapy and/or chemotherapy wasnot observed.Conclusion—Retinoblastomas and uvealmelanomas, two tumours of diVering pRBstatus, diVer also in their immunohisto-chemical pattern for markers of the G1-Sphase transition of the cell cycle. Theresults of the present study support the

concept of (a) an autoregulatory loopbetween pRB and cyclin D1 in tumourswith a functional pRB and the disruptionof this loop in the presence of pRBmutation, as well as (b) a checkpointmechanism in late G1, whose regulationvia loss of p16 or pRB, or overexpressionof cyclin D1 constitutes a common path-way to malignancy. Further, the resultsraise the possibility of cyclin D1 overex-pression having a role in the pathogenesisof uveal melanoma.(Br J Ophthalmol 1998;82:961–970)

Cyclins are cell cycle regulatory moleculeswhich complex with and activate cyclin de-pendent kinases (CDKs) to govern cell cycleprogression (Fig 1) (Table 1).1 They includecyclin D1 protein which binds with CDK4,leading to phosphorylation of the retinoblast-oma protein (pRB), a product of the tumoursuppressor gene located on chromosome13q14 which can inhibit cell proliferation(Fig 1). Phosphorylation of pRB results in achange in pRB molecule configuration, therelease of transcription factors bound to pRB,such as E2F, and ultimate cell cycle progres-sion from G1 to the S phase (Fig 1).2 3 Interestin a possible role of cyclin D1 in tumorigenesishas grown since experiments demonstratedthat induction of cyclin D1 in cells arrested inthe early G1 phase by transfection with pRBwas suYcient for cell cycle completion.4

Further, cyclin D1 overexpression has beenreported in several tumours including parathy-roid adenomas,5 mantle cell lymphoma,6 7

malignant melanoma of the skin,8 as well ascarcinomas of the breast, lung, liver, oesopha-gus, urinary bladder, head and neck, vulva, anduterine cervix.8–14

In addition to cyclin D1, other cell cycleregulatory molecules may participate in thedevelopment and progression of human tu-mours. In particular, a group of cyclindependent kinase inhibitors (CDKIs) havebeen proposed in tumorigenesis as possibletumour suppressor genes15–17 (Fig 1) (Table 1).These proteins act as negative regulatoryelements of cell proliferation by inhibiting thekinase activity of the cyclin/CDK complexesand are, thereby, implicated in the mecha-nisms of cell arrest that allow the cell time forDNA repair. Of these molecules, p16 and p21are strong candidates to participate in tumourprogression. p16 gene (CDKN2/MTS1/INK4a) is located on chromosome 9p21 andhas been found to be deleted, mutated, or

Br J Ophthalmol 1998;82:961–970 961

Department ofPathology,UniversitätsklinikumBenjamin Franklin,Freie Universität,Berlin, GermanyS E CouplandI AnagnostopoulosM HummelH Stein

Department ofOphthalmology,UniversitätsklinikumBenjamin Franklin,Freie Universität,Berlin, GermanyN BechrakisA SchülerN Bornfeld

Correspondence to:Dr S E Coupland, Institutfür Pathologie, UKBF,Hindenburgdamm 30,D-12200 Berlin, Germany.

Accepted for publication26 February 1998

downregulated by hypermethylation at highfrequency in diVerent types of tumours,including cutaneous and uvealmelanomas.15 17–20 Experimental evidence hasconfirmed p16 as an anti-oncogene and theincidence of its mutations appears to beinversely correlated with frequencies of otheroncogenic mutations, such as the loss of pRB,along the cyclin D1/CDK/p16/pRB pathway.21

p21 gene (WAF1/Cip1/SDI1), located onchromosome 6, results in G1 arrest via theinhibition of the CDKs and via an interactionwith the proliferating cell nuclear antigenprotein.22 Although it can be activated via p53independent mechanisms,23 p21 is induced by“wild type” but not mutant tumour suppressorgene p53. It is, therefore, considered to be acritical downstream eVector of p53 in cell cyclearrest.22 24 25 In normal tissues, p21 has beenassociated with maintaining growth arrest interminally diVerentiated tissues.26 Somatic mu-tations in p21 gene are rare in humantumours27; however, some have been demon-strated in human prostate cancer.28

Apart from inducing cell arrest, the tumoursuppressor gene p53, located on chromosome17q, has a second distinctive function depend-ing on the cell type. This second function is theinduction of apoptosis, also known as pro-grammed cell death (PCD).29 Wild type p53has been associated with a downregulation inthe anti-apoptotic gene bcl-2 and an upregula-tion in the bcl-2 associated gene bax, this com-bination resulting in apoptosis induction.Disruption of this balance can occur owing to aloss of normal p53 gene function or to adysregulation of BCL-2 protein expression—for example, following the characteristict(14;18) translocation observed in follicularlymphoma.30 Thereby, the balance is tippedtowards the prevention of PCD, a characteris-tic of bcl-2 which distinguishes it from otheroncogenes, which promote cell proliferation astheir mechanism of tumorigenesis. An inverse

relation between BCL-2 protein expressionand mutant p53 expression has been shown inmalignant lymphoma,31 breast32 33 and thyroidcarcinoma.34

Retinoblastoma is the most common in-traocular tumour in childhood occurring inone of 17 000 to 24 000 live births,35–38 and itmay occur either as a hereditary or a sporadictumour. The chromosome locus of the retino-blastoma gene is in the region 13q14 andaccording to the “two hit hypothesis” proposedby Knudson,39 two mutational events leadingto the inactivation or deletion of both alleles ofthe retinoblastoma locus must occur before thedevelopment of retinoblastoma. Uvealmelanoma, on the other hand, is the most fre-quent primary intraocular tumour in whiteadults with an incidence of 0.7 per 100 000.40

The pathogenesis of uveal melanoma is unclearand probably represents a multistep processinvolving the progressive and clonal accumula-tion of multiple genetic lesions aVecting proto-oncogenes and tumour suppressor genes.Recent studies have demonstrated certaingenetic changes associated with these tumours,in particular, monosomy 3, multiplication ofchromosome 8q, as well as loss of 9p.41–47

Although the role of RB gene has been clari-fied in retinoblastoma,39 little is known aboutthe cell cycle markers of G1-S phase transitionand their relation to each other in this tumour.Further, only occasional reports have ad-dressed these issues in uveal melanoma.19 Inthe present study, we investigated the expres-sion of cyclin D1 in one type of intraoculartumour where mutations of the pRB arepresent (retinoblastoma) and compared it withanother with an apparently functional pRB(uveal melanoma). In particular, we investi-gated the relation between cyclin D1 proteinexpression and that of pRB, p53, p21, and p16in these two tumours. We also analysedwhether there was a possible correlationbetween expression of these cell cycle markers

Figure 1 Interactions between cyclin D1 gene and protein (pD1) and retinoblastoma protein (pRB) as well as between the cyclin dependent kinases(CDKs) and their inhibitors (p16 and p21 via p53) during the cell cycle (modified from Lukas et al55). During early to mid G1, the concentrations of pD1are low. pRB is hypophosphorylated and has a stimulatory activity on the transcription of the cyclin D1 gene. At this time, the retinoblastoma “pockets” areoccupied by “cellular retinoblastoma binding proteins” (CRBPs), which are potent cell cycle stimulators when released later from the pRB pockets. TheCDKs are inhibited by the CDKIs, including p16 and p21, the latter being stimulated by p53. The combination of hypophosphorylated pRB, the pRBbound CRBPs, and the inhibitory eVects of the CDKIs contribute to cell cycle arrest at the restriction point (R point) in the mid to late G1 phase. Theconcentrations of pD1 increase and eventually are suYcient to combine with the CDKs, overriding the inhibitory activity of the CDKIs and resulting in(a) the displacement of the CRBPs from the RB pockets, (b) phosphorylation of the pRB, causing a change in the configuration of the pRB molecule. Thesealterations stimulate cell progression into the S phase, as well as a decrease in the transcriptional stimulus of the cyclin D1 gene. The concentration of pD1,consequentially, decreases during the S, G2, and M phases. TF-X represents a proposed transcription factor55 and BS, a possible binding site or moleculethrough which cyclin D1 indirectly interacts with pRB.

pRB

CRBPsCRBPs

pRB pRB

pD1p16, p53 via p21

CDKs

CDKs

pD1

BS

pD1 pD1

++

+

+

++

TF-X TF-X TF-X

Genepromoter

Cyclin D1 gene Genepromoter

Genepromoter

Cyclin D1 gene Cyclin D1 gene

Restriction point

pD1

CDKsCRBPs

P PP

P

PP

p16, p53 via p21 p16, p53 via p21

Early–mid G1 (0–6 hours) Mid–late G1 (6–12 hours) G1/S, S, G2, M (12–24 hours)

962 Coupland, Bechrakis, Schüler, et al

of G1-S phase transition and the growthfraction of the tumour, as determined by theKi-67 antigen, or with BCL-2 expression.

MethodsCONVENTIONAL HISTOLOGY AND

IMMUNOHISTOLOGY

Ninety eyes were enucleated for either uvealmelanoma (n=66) or retinoblastoma (n=24).Eight eyes donated for corneal transplantationwith an average postmortem time of 12 hoursserved as controls. All enucleated eyes werefixed in 10% buVered formalin and embeddedin paraYn wax. Conventional histologicalstains of the uveal melanomas were assessed forcell type using the modified Callendersystem48—that is, spindle A, spindle B, epithe-lioid, and mixed cell type. Retinoblastomaswere graded as to the degree of diVerentiationaccording to a system proposed by Nork andcoworkers.49 Briefly, the grades of diVerenti-ated were: (1) poorly diVerentiated—cells withhigh nuclear to cytoplasmic ratios with a highmitotic index. Pseudorosettes and Homer–Wright rosettes were found in such areas; (2)moderately diVerentiated—cells with moderatenuclear to cytoplasmic ratios, a moderatemitotic index, and possible pseudorosettes andFlexner–Winterstein rosettes; (3) welldiVerentiated—cells with a low nuclear to cyto-plasmic ratio, a low mitotic index, and the for-mation of florets. The recently described“zones 1, 2, and 3” around pseudorosettes49

were used for the evaluation of tumour cellstaining. Briefly, these were as follows “zone 1”(central region near blood vessel), “zone 2”(adjacent to edge of the pseudorosette), and“zone 3” (edge of pseudorosette).49

Additional slides were stained for immuno-histochemical studies using several monoclonaland polyclonal antibodies that are reactive inparaYn embedded tissues. An antigen retrievalmethod using a pressure cooker was performedbefore immunohistochemical staining.50 Thestaining consisted of a first stage incubationwith the following primary monoclonal anti-bodies: cyclin D1 (clone P2D11F11; Novocas-tra; Germany); retinoblastoma protein (pRB)(clone G3-245 which binds to an epitopebetween amino acids 300–380 of human RB;Pharmingen; Germany); p53 (clone DO7which recognises both wild type and mutantp53 proteins; Dako; Denmark); p21 (cloneDCS-60.2; Neomarkers; Germany); p16(clone DCS-50.1/A7; Neomarkers; Germany);BCL-2 (clone 124; Dako; Denmark); MIB-1(antigen Ki-67 which reacts with a DNA asso-ciated antigen in the nuclei at all phases of thecell cycle except the resting phase51; theantibody was kindly provided by Dr J Gerdes,

Borstel, Germany); and glial fibrillary acidprotein (clone 6F2; GFAP; Dako; Denmark).The antibodies were made visible with an indi-rect immunoperoxidase method for p53,whereas the alkaline phosphatase anti-alkalinephosphatase (APAAP) method was used todemonstrate the binding of the remainingantibodies.52 In heavily pigmented tumours,the sections were placed in hydrogen peroxidefor 18 hours to remove melanin pigmentationfrom the tumour cells before cover slipping theslides, as previously described.53

Cells were considered positive for cyclin D1,pRB, p53, p16, p21, and for MIB-1 only whendistinct nuclear staining was identified. Thepercentage of immunoreactive nuclei in theuveal melanoma and retinoblastoma wasevaluated by counting at least 5 × 100 cellsusing the 40× objective (Olympus, BH2). Posi-tive controls included cases of mantle celllymphoma for cyclin D1, colon carcinoma forp21, and gastric and breast carcinoma for p16.Negative controls were obtained by omittingthe primary monoclonal antibodies.

STATISTICAL ANALYSIS

Comparison of diVerences was performedusing the Student’s t test. p Values at less than0.05 were interpreted to be statistically signifi-cant.

ResultsCLINICAL FEATURES

The patients (n= 23) with retinoblastoma con-sisted of eight females and 15 males with anage range of 4 months to 19 years; mean 3years. Twelve patients had bilateral tumoursand 11 unilateral. Sixteen patients were treatedwith primary enucleation; five with chemore-duction before planned enucleation; two withcombined chemotherapy and radiotherapybefore enucleation; and one patient with radio-therapy alone.

The patients (n=66) with uveal melanomaconsisted of 39 females and 27 males with anage range of 9–88 years; mean 59 years. Fiftyfive patients had been treated with primaryenucleation or local excision of the tumour;nine with radiotherapy before enucleation.

CONVENTIONAL AND IMMUNOHISTOLOGY

The results of the immunohistochemical inves-tigations are summarised in Table 2.

Normal choroid and retinaCyclin D1 was observed in occasional ganglioncells of the normal retina (Fig 2a); a similarstaining pattern was observed for p53 (Fig 2b)and p16. Positive staining for pRB and p21(Fig 2c) was observed in all cell layers of thenormal retina. BCL-2 positivity was observedin scattered ganglion cells, the bipolar cells,and in perivascular astrocytes (Table 2) (Fig2d).

The choroidal melanocytes were negative forcyclin D1, pRB, p53, and p16. Occasionalmelanocytes, as well as scattered reactive T andB lymphocytes, were positive for p21 andBCL-2 (Fig 2e) (Table 2).

Table 1 Molecules involved in executing cell cycle arrest atthe restriction point (R point) between G1-S phase (A)and those involved in recommencement of cell cycling afterthe R point (B)

A B

CDKIs: p53 via p21; p16 Cyclin D1hypophosphorylated pRB CDKs

hyperphosphorylated pRBCRBPs (for example, E2F)

Cyclin D1 and intraocular tumours 963

RetinoblastomaCyclin D1 positive cells were rarely seen inretinoblastoma and when present correspondedto glial cells (Fig 3a), also positive for GFAP.Similarly, retinoblastoma cells were negative for

pRB using the applied antibody, staining forpRB being observed only in vascular endothe-lial cells and some perivascular GFAP positiveglial cells (Table 2). Excluding those retinoblas-tomas which had been treated with chemo-

Table 2 Summary of the immunohistochemical findings for cyclin D1 and related G1-S phase proteins in the normalretina, untreated retinoblastoma, and untreated choroidal melanoma

Normal retina Retinoblastoma Normal choroid Choroidal melanoma

Cyclin D1 GCL positive negative negative positive (2–75%)*pRB all layers positive negative negative positive (5–65%)p53 GCL positive PDA>MDA>WDA negative positive (2–30%)p21 all layers positive PDA and MDA lymphocytes only positive (5–30%)p16 GCL positive PDA and MDA negative positive (5–25%)BCL-2 GCL, BC and PA positive GC lymphocytes only positive (100%)MIB-1 NA PDA (80%) NA 5–50%

MDA (30%)WDA (5%)

pRB = retinoblastoma protein; GCL = ganglion cell layer; BC = bipolar cells; PA = perivascular astrocytes; GC = glial cells; PDA= poorly diVerentiated area; MDA = moderately diVerentiated area; WDA = well diVerentiated area. NA = not applicable.*Percentages in parentheses represent the number of tumour cells per 100 positive for the various markers.

Figure 2 Normal retina with (a) cyclin D1 staining in occasionalganglion cells (arrow) (×20 objective); (b) p53 positivity in scatteredganglion cells (×40 objective); (c) granular p21 staining in all cell layers ofthe retina (×20 and ×40 objective); (d) BCL-2 positivity in glial cells and,at the higher magnification, in occasional ganglion cells (×20 and ×40objective). (e) Normal choroid with scattered BCL-2 positive reactivelymphocytes (arrow) (×40 objective); occasional melanocytes demonstratedweak BCL-2 positivity following bleaching (not shown).

964 Coupland, Bechrakis, Schüler, et al

therapy, radiation, or with a combination of thetwo before enucleation, most retinoblastomasdemonstrated all three grades of cell diVerentia-tion. These were reflected to some extent by thegrowth fraction of the tumour cells as well as bythe percentages of cells positive for the p53,p21, and p16. The highest proliferating cellpopulations were observed in the poorly diVer-entiated areas with growth fractions of 60–90%(mean 80%). In the moderately and well diVer-entiated areas, the growth fractions weresignificantly lower, averaging 30% and 5%,respectively (Fig 3b). Highest percentages ofp53 positive cells (mean 60%) were observed inthe poorly diVerentiated areas; lower percent-ages (mean 25%) in the moderately diVerenti-

ated areas within the Flexner–Wintersteinrosettes; and only scattered p53 positive cellswere observed in the well diVerentiated areas.p21 positive tumour cells were observed in thepoorly and moderately diVerentiated areas;occasional cells were positive for p21 in the welldiVerentiated tumour areas (Table 2). Therecently described zonal diVerentiation of p53and p21 staining cells in pseudorosettes—p21adjacent to the central blood vessel (zone 1)and p53 adjacent to the pseudorosette edge(zone 2)49—was observed in many retinoblasto-mas. A large proportion of pseudorosettes,however, demonstrated p53 and p21 positivecells in both zones (Fig 3c and d, respectively).p16 stained positively in high percentages in

Figure 3 (a) Cyclin D1 positivity in reactive glial cells (arrow) within retinoblastoma; the tumour cells are negative (×20 objective). (b) The degree ofdiVerentiation within the retinoblastomas is reflected by the corresponding MIB-1 growth fractions (×20 objective); (c) p53; and (d) p21 positivitysurrounding a pseudorosette where staining of these markers is present in all “zones” (×40 objective); (e) p16 staining in poorly and moderatelydiVerentiated tumours areas within a retinoblastoma (×20 objective). (f) Perivascular glial cells positive for BCL-2 in a retinoblastoma; the tumour cells arenegative (×40 objective).

Cyclin D1 and intraocular tumours 965

areas of poor and moderate diVerentiation (Fig3e). Cells in zones 1 and 2 of the pseudorosetteswere positive for p16. Only rare p16 positivecells were observed in well diVerentiatedtumour areas.

Perivascular glial cells in areas of all gradesof cellular diVerentiation were positive forBCL-2 (Fig 3f). Further, scattered or occa-sional groups of glial cells positive for BCL-2were present between the tumour cells.

The nine retinoblastomas from patientstreated with chemotherapy, chemoreduction,or radiation before enucleation demonstratedextensive calcification of the devitalised neo-

plastic cells, cystic degeneration, as well asreactive gliosis. Changes to retinoblastoma fol-lowing treatment as well as those observed inadjacent ocular tissue have been described inmore detail elsewhere.54 Three tumours dem-onstrated vital tumour with growth fractions of10% in one case and 90% in the other two.Occasional vital retinoblastoma cells werepositive for p53, p21, and p16; there was noalteration in the expression of these markerscompared with the “untreated” tumours in thevital areas of the tumour. All post-treatmentretinoblastomas were negative for cyclin D1,pRB, and BCL-2.

Figure 4 Cyclin D1 positivity in a spindle (a) and epithelioid (b) choroidal melanoma (×20 objective). pRB positivity ina spindle (c) and epithelioid (d) choroidal melanoma (×20 objective). (e) Clear p53 positivity in choroidal melanoma(arrow) (×20 objective). (f) p21 staining in occasional tumour cells of a mixed cell choroidal melanoma (×20 objective).(g) p16 positivity in scattered tumour cells of a mixed cell choroidal melanoma (×40 objective). (h) BCL-2 staining in aspindle choroidal melanoma (×20 objective).

966 Coupland, Bechrakis, Schüler, et al

A correlation between the above G1-S cellcycle markers and the laterality of the tumourswas not demonstrated.

Uveal melanomasForty five of the uveal melanomas arose fromthe choroid; 18 from the ciliary body; and onein the iris. Of the choroidal melanomas, 37were posterior to, or close to the equator; eighttumours were anterior to this with no infiltra-tion of the pars plana of the ciliary body. Theuveal melanomas consisted of 21 spindle B, 29mixed tumours with spindle B predominance,seven pure epithelioid tumours, and sevenmixed tumours with epithelioid predominance.Extraocular extension of tumour was observedin nine cases.

Excluding those uveal melanoma which hadbeen treated with radiotherapy before enuclea-tion, nuclear positivity of cyclin D1 wasobserved in 85% of cases with the percentageof positive cells ranging between 2% and 75%(Table 2). The cyclin D1 percentage expres-sion correlated positively with the growth frac-tion of the tumours which varied between 5%and 55% (p<0.05, Student’s t test). Thepercentage of cells positive for MIB-1 and cyc-lin D1 significantly correlated with cell type:pure or mixed epithelioid tumours demon-strated a greater percentage of MIB-1 and cyc-lin D1 positive cells than spindle cell tumours(p<0.05, respectively) (Fig 4a and b). Further,the percentage of cyclin D1 positive cells wasgreater in the anteriorly located tumours orthose which showed extraocular tumour exten-sion (p<0.05).

pRB tumour cell positivity varied between5% and 65% (Table 2) and a significant corre-lation was observed between the percentage ofcells expressing pRB and the MIB-1 growthfraction of the tumour (p<0.05). Further, apositive correlation existed between pRB andthe cyclin D1 expression, the cell type (less inspindle cell tumours) (Fig 4c and d) andanatomical location of the tumour (higher per-centage in the anteriorly located tumours)(p<0.05, respectively).

p53 expression was found in 60% of cases,with the percentage of positive cells varyingbetween 2% and 30% (Fig 4e) (Table 2).Higher percentages of p53 positivity tended tooccur in epithelioid cells; however, this was notstatistically significant (p=0.09). Further, sta-tistically significant correlations between p53expression and the MIB-1 growth fraction,cyclin D1 and pRB positivity, and the tumourlocation were not observed.

Occasional tumour cells were positive forp21 in most cases (65%), range 5–30% (mean8%) (Fig 4f). A statistically significant inverserelation between p53 and p21 was observedwith a decreased p21 expression in tumourswith high percentage p53 expression (p<0.05).Double staining demonstrated that p53 andp21 positive cells were distinct from oneanother (not shown). A significant correlationbetween p21 positive cells and the MIB-1growth fraction, cyclin D1 and pRB positivity,tumour cell type or tumour location was notobserved.

Immunohistochemical detection with p16showed both nuclear and cytoplasmic stainingin uveal melanoma; p16 cells were thescattered cells seen in 45% of cases and thispositivity varied between 5% and 25% (mean12%) of tumour cells (Table 2) (Fig 4g). Therewas no correlation with cell type, tumour loca-tion, or the percentage of cells positive for pRB.Double staining demonstrated exclusive stain-ing for p16 and pRB in uveal melanoma.

All uveal melanomas stained for BCL-2,with no variation being observed betweentumour location, tumour cell type (Fig 4h),MIB-1 growth fraction, cyclin D1, pRB, p53,p21, or p16 positivity. Despite bleachingprocedures, slight variations in the intensity ofthe BCL-2 staining occurred in heavily pig-mented tumours owing to the cytoplasmiclocation of the BCL-2 protein.

The nine uveal melanomas which had beentreated with radiotherapy before enucleationdemonstrated necrosis as well as a massiveinfiltration of melanomacrophages. Thegrowth fraction of these tumours was signifi-cantly lower in comparison with untreatedtumours, varying between 1% and 5%, mean3% (p<0.05). Vital melanoma cells were posi-tive for BCL-2 and, occasionally, for cyclin D1,pRB, p53, p21, and p16. There was no changein expression of these markers in the vitaltumour areas compared with the “untreated”tumours.

DiscussionThe present study represents the first immuno-histochemical investigation and description ofcyclin D1 protein expression in retinoblastomaand uveal melanoma. The two tumours werechosen for their diVering retinoblastoma genestatus in order to investigate the expression ofcyclin D1 and related G1-S transition proteinsdependent on pRB status. Recent studies oncell lines and data from in vitro systems haveprovided evidence for an autoregulatory loopbetween cyclin D1 and pRB in normal cellsand tumour cells with a functional pRB(Fig 1).55 According to this model, transcrip-tion of cyclin D1 gene is stimulated byhypophosphorylated pRB in the early stages ofG1 and is later repressed by its hyperphospho-rylated form (Fig 1). Disruption of this loophas been proposed to occur in three ways. Thefirst is the binding of DNA viruses to the pRBbinding “pockets”55 and will not be furtheraddressed here. The second is the mutation ofthe pRB molecule, which predominantly in-volves the pRB binding pockets.55 In humantumours with mutations in both alleles of theRB gene, functional pRB is absent and can nolonger regulate transcriptional factors such asE2F. E2F is, thereby, freed from the pRB bind-ing pockets and from the inhibitory regulationof pRB at inappropriate times and, conse-quently, cell proliferation is stimulated. Fur-ther, there is a decrease in the transcriptionalstimulus of cyclin D1 by the mutated pRB.55

The third mechanism, by which the autoregu-latory loop between pRB and cyclin D1 can bedisrupted, is the overexpression of cyclin D1 bygene amplification, resulting in accelerated

Cyclin D1 and intraocular tumours 967

phosphorylation of pRB and the inactivation ofits inhibitory eVect on cell cycle progression.55

The exact mechanism by which cyclin D1overexpression has its oncogenic eVect isunclear. Two mechanisms have beenproposed—by directly binding to pRB or byreducing the levels pRB in the cell.2 4 Bothmechanisms have been recently questioned instudies on low grade B cell lymphomas.7

In the current study, retinoblastoma tumourcells were negative for cyclin D1 and pRBusing the antibodies applied. Uveal melanomacells, in contrast, demonstrated positivity forboth cyclin D1 and pRB in varying percent-ages, with a positive correlation existingbetween these two markers (p<0.05), as well aswith the growth fraction of the choroidalmelanoma (p<0.05). Our data are in agree-ment with those of previous studies investigat-ing tissue sections for pRB56 57 and cell lines forcyclin D1.55 The absence of pRB and cyclin D1protein staining in retinoblastoma cells may beexplained by the above mentioned secondmechanism leading to disruption of the au-toregulatory loop between cyclin D1 and pRB.The cause of cyclin D1 overexpression in theuveal melanomas—for example, via a heteroge-neous cyclin D1 gene promoter or via genetranslocation, as in the case of mantle celllymphoma, must be investigated further. Thepositive relation between cyclin D1 and pRBobserved in choroidal melanoma, however,would tend to support the suggestion ofZukerberg and colleagues that cyclin D1 doesnot exerts its tumorigenic eVect by pRB levelreduction.7

Strong expression of p16, a protein whichacts as a kinase inhibitor of the cyclin/CDKcomplexes, was observed in retinoblastoma inareas of moderate and poor diVerentiation. Itwas also demonstrated in pRB negative cells in45% of uveal melanomas. The variable expres-sion of p16 observed in uveal melanomas in thecurrent study parallels to some extent the find-ings of Ohta et al who reported chromosomalabnormalities involving the p16 gene in 32% ofcases examined.19 In general, the results of p16expression in both tumour types are consistentwith studies which have demonstrated highlevels of p16 expression in cell lines with lack ofpRB function58 and would support the pro-posal of a negative feedback system betweenthe two proteins.59 60 Further, they are in agree-ment with the results of immunohistochemicalinvestigations of other tumours where areciprocal expression of p16 and pRB wasdemonstrated.61 62 These current models donot explain, however, the absence of p16expression in well diVerentiated areas ofretinoblastomas (also negative for pRB). To-gether with biochemical analysis further inves-tigations are presently being conducted withnewer p16 antibodies, in order to underline thesignificance of the present findings.

Some variation exists between the previouslyreported investigations of p53 expression inuveal melanoma63 64 and can most likely beaccounted for by the diVering fixation tech-niques, the diVering p53 antibodies, as well asdiVering antigen retrieval techniques used.

Such problems with p53 immunohistochemi-cal investigations have been recently addressedin other human cancers in detail and the valueof the immunohistochemical analysis alone asbeing indicative of mutations in the p53 genehas been questioned.65 In the current study, themajority (60%) of uveal melanomas demon-strated p53 positivity with cell percentagesvarying between 2% and 30%. Previousauthors have reported 54%63 and 67%64 ofcases of uveal melanomas being positive forp53. Details regarding the percentage oftumour cells positive for this marker were notgiven in the latter study. One group of investi-gators defined p53 overexpression as beinggreater than 10% and this was only observed infive patients, all of whom had been treated withpreoperative telecobalt or ruthenium irradia-tion before enucleation.63 Although our resultsdo not support such an association betweenp53 expression and irradiation, we agree withthese authors’ observation that p53 positivitytended to be greater in epithelioid tumours.

The critical downstream eVector of p53 spe-cific pathway growth control, p21, was ob-served in both intraocular tumours in the cur-rent investigation. Further, the inverse relationbetween p53 and p21 protein expressiondescribed in normal tissue and in othertumours,26 could be demonstrated in somepseudorosettes within the retinoblastomas andin uveal melanomas. Nork and coworkers haverecently proposed a zonal arrangement of p53and p21 positive cells in tumour areas adjacentto pseudorosettes in retinoblastoma.49 Theauthors reported that zone 1, 2, and 3 cellssurrounding the pseudorosettes have the high-est staining of cells for p21, p53, andtransferase mediated biotin dUTP nick endlabelling, respectively and that an overexpres-sion of p53 in zone 2 leads to the ultimateonset of apoptosis observed at the edge of thepseudorosette.49 Although a similar pattern forp53 and p21 staining was observed in somepseudorosettes in our tumours, such a strictlydefined distribution of these two cell markerswas not observed in all pseudorosettes and inall tumours. In the choroidal melanomas, p21positivity decreased in tumours with a highpercentage p53 expression. It would be ex-pected from these results that, as a result of adecrease in the “arrest” function of p21, apositive correlation may exist between p53expression and the MIB-1 growth fraction.Although such a correlation between p53 andproliferation markers has been observed inimmunohistochemical investigations of othertumours,66–68 such a relation could not be dem-onstrated in this study. The only influenceobservable on the growth fraction of uvealmelanomas was that of irradiation whichresulted in a significant decrease in the MIB-1growth fraction, as previously reported byother authors.69–71

Apart from its cell arrest inducing function,p53 is also involved in the induction of PCD; adirect correlation between p53 expression andapoptosis has been demonstrated in sometumours.34 Retinoblastoma cells were negativefor BCL-2, the antidote to PCD; perivascular

968 Coupland, Bechrakis, Schüler, et al

glial cells and scattered astrocytes within thetumour islands only were positive for BCL-2.This absence of BCL-2, together with theextensive apoptosis observed in the majority ofthe retinoblastomas, would tend to support thesuggestion of Nork et al, that apoptosis inretinoblastoma is most likely p53 induced.49

Our findings, however, contrast with those ofYuge et al, who described weak positivity forBCL-2 in all retinoblastoma cells.57 Thereasons for the discrepancy between results arenot clear, although they are likely to beexplained by the diVering preparation of theslides before immunohistochemistry (hydro-gen peroxide incubation; microwave antigenretrieval) as well as in the diVering immunohis-tochemical techniques.57

Uveal melanoma cells, in contrast, were con-sistently positive for BCL-2, in agreement withprevious results where BCL-2 positivity hasbeen described in 100%72 and 90%73 of cases.BCL-2 is considered to have a specific role inthe survival of melanocytes.74 Although a posi-tive correlation between the oncogene c-mycand BCL-2 has been established in choroidalmelanomas,73 there was no significant correla-tion between BCL-2 and any of the G1-Sphase markers investigated in the presentstudy. The interpretation of BCL-2 staining inuveal melanoma most likely requires the addi-tional results of immunohistochemical stainingpatterns of other members of the BCL-2 fam-ily, such as BAX, which have to date beenhampered by the unsuitability of available anti-bodies for formalin fixed paraYn embeddedtissue sections.

In conclusion, retinoblastomas and uvealmelanomas diVer in their immunohistochemi-cal pattern for cell cycle markers of the G1-Sphase transition. Cyclin D1 protein expressionis present in most uveal melanomas and absentin retinoblastomas, in agreement with therecently proposed autoregulatory feedbackloop model proposed by Lukas andcoworkers.55 A positive correlation exists be-tween cyclin D1 cell positivity and tumour celltype, location, growth fraction, and pRBpositivity in uveal melanoma. The resultspresented are based on immunohistochemicalinvestigations only and require molecular bio-logical studies to underline their significance.The presence of cyclin D1 overexpression inchoroidal melanoma suggests that othermechanisms at the molecular level, such as apossible translocation or heterogeneous pro-moter of the cyclin D1 gene, may be involvedin their tumorigenesis. It also remains to bedetermined whether the expression of theseG1-S phase transition proteins, particularly ofcyclin D1 protein, has a role in determiningtumour behaviour—for example, the onset oftumour metastasis.

The authors thank Mrs Helga Zimmerman-HöVken and MrsHeidrun Protz for their superb technical assistance.

1 Sherr CJ. Mammalian G1 cyclins. Cell 1993;73:1059–65.2 Dowdy SF, Hinds PW, Louie K, et al. Physical interaction of

the retinoblastoma protein with human D cyclins. Cell1993;73:499–511.

3 Ewen ME, Sluss HK, Sherr CJ, et al. Functional interactionsof the retinoblastoma protein with mammalian D-type cyc-lins. Cell 1993;73:487–97.

4 Hinds PW, Mittnacht S, Dulic V, et al. Regulation of retino-blastoma protein functions by ectopic expression of humancyclins. Cell 1992;70:993–1006.

5 Motokura T, Bloom T, Kim HG, et al. A novel cyclinencoded by a bcl1-linked candidate oncogene [seecomments]. Nature 1991;350:512–5.

6 Ott MM, Helbing A, Ott G, et al. bcl-1 rearrangement andcyclin D1 protein expression in mantle cell lymphoma. JPathol 1996;179:238–42.

7 Zukerberg LR, Benedict WF, Arnold A, et al. Expression ofthe retinoblastoma protein in low-grade B-cell lymphoma:relationship to cyclin D1. Blood 1996;88:268–76.

8 Bartkova J, Lukas J, Strauss M, et al. Cyclin D1 oncoproteinaberrantly accumulates in malignancies of diverse his-togenesis. Oncogene 1995;10:775–8.

9 Bartkova J, Lukas J, Muller H, et al. Cyclin D1 proteinexpression and function in human breast cancer. Int J Can-cer 1994;57:353–61.

10 Buckley MF, Sweeney KJ, Hamilton JA, et al. Expressionand amplification of cyclin genes in human breast cancer.Oncogene 1993;8:2127–33.

11 Kurzrock R, Ku S, Talpaz M. Abnormalities in the PRAD1(CYCLIN D1/BCL-1) oncogene are frequent in cervicaland vulvar squamous cell carcinoma cell lines. Cancer1995;75:584–90.

12 Mate JL, Ariza A, Aracil C, et al. Cyclin D1 overexpressionin non-small cell lung carcinoma: correlation with Ki67labelling index and poor cytoplasmic diVerentiation. JPathol 1996;180:395–9.

13 Michalides R, van Veelen N, Hart A, et al. Overexpression ofcyclin D1 correlates with recurrence in a group offorty-seven operable squamous cell carcinomas of the headand neck. Cancer Res 1995;55:975–8.

14 Naitoh H, Shibata J, Kawaguchi A, et al. Overexpression andlocalization of cyclin D1 mRNA and antigen in esophagealcancer. Am J Pathol 1995;146:1161–9.

15 Hirama T, KoeZer HP. Role of the cyclin-dependent kinaseinhibitors in the development of cancer. Blood 1995;86:841–54.

16 Hunter T, Pines J. Cyclins and cancer. Cell 1991;66:1071–4.17 Hunter T, Pines J. Cyclins and cancer. II: Cyclin D and

CDK inhibitors come of age [see comments]. Cell 1994;79:573–82.

18 Herman JG, Merlo A, Mao L, et al. Inactivation of theCDKN2/p16/MTS1 gene is frequently associated withaberrant DNA methylation in all common human cancers.Cancer Res 1995;55:4525–30.

19 Ohta M, Berd D, Shimizu M, et al. Deletion mapping ofchromosome region 9p21-p22 surrounding the CDKN2locus in melanoma. Int J Cancer 1996;65:762–7.

20 Okamoto A, Demetrick DJ, Spillare EA, et al. Mutationsand altered expression of p16INK4 in human cancer. ProcNatl Acad Sci USA 1994;91:11045–9.

21 Otterson GA, Khleif SN, Chen W, et al. CDKN2 genesilencing in lung cancer by DNA hypermethylation andkinetics of p16INK4 protein induction by 5-aza 2’deoxycy-tidine. Oncogene 1995;11:1211–6.

22 el Deiry WS, Tokino T, Velculescu VE, et al. WAF1, apotential mediator of p53 tumor suppression. Cell 1993;75:817–25.

23 Michieli P, Chedid M, Lin D, et al. Induction ofWAF1/CIP1 by a p53-independent pathway. Cancer Res1994;54:3391–5.

24 Harper JW, Adami GR, Wei N, et al. The p21 Cdk-interacting protein Cip1 is a potent inhibitor of G1 cyclin-dependent kinases. Cell 1993;75:805–16.

25 Harper JW, Elledge SJ, Keyomarsi K, et al. Inhibition ofcyclin-dependent kinases by p21. Mol Biol Cell 1995;6:387–400.

26 Doglioni C, Pelosio P, Laurino L, et al. p21/WAF1/CIP1expression in normal mucosa and in adenomas and adeno-carcinomas of the colon: its relationship with diVerentia-tion. J Pathol 1996;179:248–53.

27 Shiohara M, el Deiry WS, Wada M, et al. Absence of WAF1mutations in a variety of human malignancies. Blood 1994;84:3781–4.

28 Gao X, Chen YQ, Wu N, et al. Somatic mutations of theWAF1/CIP1 gene in human prostate cancer. Oncogene1995;11:1395–8.

29 Prokocimer M, Rotter V. Structure and function of p53 innormal cells and their aberrations in cancer cells:projectionon the hematologic cell lineages. Blood 1994;84:2391–411.

30 Yunis JJ. Specific fine chromosomal defects in cancer: anoverview. Hum Pathol 1981;12:503–15

31 Piris MA, Villuendas R, Martinez JC, et al. p53 expressionin non-Hodgkin’s lymphomas: a marker of p53 inactiva-tion? Leuk Lymphoma 1995;17:35–42.

32 Haldar S, Negrini M, Monne M, et al. Down-regulation ofbcl-2 by p53 in breast cancer cells. Cancer Res 1994;54:2095–7.

33 Silvestrini R, Veneroni S, Daidone MG, et al. The Bcl-2protein: a prognostic indicator strongly related to p53 pro-tein in lymph node-negative breast cancer patients. J NatlCancer Inst 1994;86:499–504.

34 Basolo F, Pollina L, Fontanini G, et al. Apoptosis and prolif-eration in thyroid carcinoma: correlation with bcl-2 andp53 protein expression. Br J Cancer 1997;75:537–41.

35 Kock E, Naeser P. Retinoblastoma in Sweden 1958–1971. Aclinical and histopathological study. Acta OphthalmolCopenh 1979;57:344–50.

Cyclin D1 and intraocular tumours 969

36 Mahoney MC, Burnett WS, Majerovics A, et al. The epide-miology of ophthalmic malignancies in New York State.Ophthalmology 1990;97:1143–7.

37 Pendergrass TW, Davis S. Incidence of retinoblastoma inthe United States. Arch Ophthalmol 1980;98:1204–10.

38 Sanders BM, Draper GJ, Kingston JE. Retinoblastoma inGreat Britain 1969–80: incidence, treatment, and survival.Br J Ophthalmol 1988;72:576–83.

39 Knudson AG Jr. The genetics of childhood cancer. Cancer1975;35:1022–6.

40 Jensen OA. Malignant melanomas of the human uvea:25-year follow-up of cases in Denmark, 1943–1952. ActaOphthalmol Copenh 1982;60:161–82.

41 Horsman DE, Sroka H, Rootman J, et al. Monosomy 3 andisochromosome 8q in a uveal melanoma. Cancer GenetCytogenet 1990;45:249–53.

42 Horsman DE, White VA. Cytogenetic analysis of uvealmelanoma. Consistent occurrence of monosomy 3 and tri-somy 8q. Cancer 1993;71:811–9.

43 Prescher G, Bornfeld N, Becher R. Nonrandom chromo-somal abnormalities in primary uveal melanoma. J NatlCancer Inst 1990;82:1765–9.

44 Prescher G, Bornfeld N, Horsthemke B, et al. Chromosomalaberrations defining uveal melanoma of poor prognosis[letter]. Lancet 1992;339:691–2.

45 Sisley K, Rennie IG, Cottam DW, et al. Cytogenetic findingsin six posterior uveal melanomas:involvement of chromo-somes 3, 6, and 8. Genes Chromosom Cancer 1990;2:205–9.

46 Sisley K, Cottam DW, Rennie IG, et al. Non-random abnor-malities of chromosomes 3, 6, and 8 associated with poste-rior uveal melanoma. Genes Chromosomes Cancer 1992;5:197–200.

47 Speicher MR, Prescher G, du Manoir S, et al. Chromosomalgains and losses in uveal melanomas detected by compara-tive genomic hybridization. Cancer Res 1994;54:3817–23.

48 McLean IW, Foster WD, Zimmerman LE, et al. Modifica-tions of Callender’s classification of uveal melanoma at theArmed Forces Institute of Pathology. Am J Ophthalmol1983;96:502–9.

49 Nork TM, Poulsen GL, Millecchia LL, et al. p53 regulatesapoptosis in human retinoblastoma. Arch Ophthalmol1997;115:213–9.

50 Norton AJ, Jordan S, Yeomans P. Brief, high-temperatureheat denaturation (pressure cooking): a simple andeVective method of antigen retrieval for routinely processedtissues. J Pathol 1994;173:371–9.

51 Gerdes J, Lemke H, Baisch H, et al. Cell cycle analysis of acell proliferation-associated human nuclear antigen definedby the monoclonal antibody Ki-67. J Immunol 1984;133:1710–5.

52 Cordell JL, Falini B, Erber WN, et al. Immunoenzymaticlabeling of monoclonal antibodies using immune com-plexes of alkaline phosphatase and monoclonal anti-alkaline phosphatase (APAAP complexes). J HistochemCytochem 1984;32:219–29.

53 Fuchs U, Kivela T, Summanen P, et al. An immunohisto-chemical and prognostic analysis of cytokeratin expressionin malignant uveal melanoma. Am J Pathol 1992;141:169–81.

54 Bechrakis NE, Bornfeld N, Schüler A, et al. Clinicopatho-logical features of retinoblastoma after primary chemore-duction. Arch Ophthalmol 1998;(in press).

55 Lukas J, Muller H, Bartkova J, et al. DNA tumor virusoncoproteins and retinoblastoma gene mutations share the

ability to relieve the cell’s requirement for cyclin D1 func-tion in G1. J Cell Biol 1994;125:625–38.

56 Nork TM, Schwartz TL, Doshi HM, et al. Retinoblastoma.Cell of origin. Arch Ophthalmol 1995;113:791–802.

57 Yuge K, Nakajima M, Uemura Y, et al. Immunohistochemi-cal features of the human retina and retinoblastoma.Virchows Arch 1995;426:571–5.

58 Otterson GA, Kratzke RA, Coxon A, et al. Absence ofp16INK4 protein is restricted to the subset of lung cancerlines that retains wildtype RB. Oncogene 1994;9:3375–8.

59 Yeager T, Stadler W, Belair C, et al. Increased p16 levels cor-relate with pRb alterations in human urothelial cells. Can-cer Res 1995;55:493–7.

60 Beltinger CP, White PS, Sulman EP, et al. No CDKN2mutations in neuroblastomas. Cancer Res 1995;55:2053–5.

61 Sakaguchi M, Fujii Y, Hirabayashi H, et al. Inversely corre-lated expression of p16 and Rb protein in non-small celllung cancers: an immunohistochemical study. Int J Cancer1996;65:442–5.

62 Tsuzuki T, Tsunoda S, Sakaki T, et al. Alterations of retino-blastoma, p53, p16(CDKN2), and p15 genes in humanastrocytomas. Cancer 1996;78:287–93.

63 Janssen K, Kuntze J, Busse H, et al. p53 oncoprotein overex-pression in choroidal melanoma. Mod Pathol 1996;9:267–72.

64 Tobal K, Warren W, Cooper CS, et al. Increased expressionand mutation of p53 in choroidal melanoma. Br J Cancer1992;66:900–4.

65 Wynford Thomas D. P53 in tumour pathology: can we trustimmunocytochemistry? [Editorial] [see comments]. JPathol 1992;166:329–30.

66 Costa MJ, Walls J, Ames P, et al. Transformation inrecurrent ovarian granulosa cell tumors:Ki67 (MIB-1) andp53 immunohistochemistry demonstrates a possible mo-lecular basis for the poor histopathologic prediction ofclinical behavior. Hum Pathol 1996;27:274–81.

67 Golusinski W, Olofsson J, Szmeja Z, et al. p53, PCNA andKi67 in laryngeal cancer. Pol J Pathol 1996;47:175–82.

68 Nawa G, Ueda T, Mori S, et al. Prognostic significance ofKi67 (MIB1) proliferation index and p53 over-expressionin chondrosarcomas. Int J Cancer 1996;69:86–91.

69 Char DH, Huhta K, Waldman F. DNA cell cycle studies inuveal melanoma. Am J Ophthalmol 1989;107:65–72.

70 Mooy CM, de Jong PT, Van der Kwast TH, et al. Ki-67immunostaining in uveal melanoma. The eVect of pre-enucleation radiotherapy. Ophthalmology 1990;97:1275–80.

71 Schilling H, Weng Sehu K, Lee WR. A histologic study(including DNA quantification and Ki-67 labeling index)in uveal melanomas after brachytherapy with rutheniumplaques. Invest Ophthalmol Vis Sci 1997;38:2081–92.

72 Jay V, Yi Q, Hunter WS, Zielenska M. Expression of bcl-2 inuveal malignant melanoma. Arch Pathol Lab Med 1996;120:497–8.

73 Mooy CM, Luyten GP, de Jong PT, et al. Immunohisto-chemical and prognostic analysis of apoptosis and prolif-eration in uveal melanoma. Am J Pathol 1995;147:1097–104.

74 Yang E, Korsmeyer SJ. Molecular thanatopsis:a discourseon the BCL2 family and cell death. Blood 1996;88:386–401.

970 Coupland, Bechrakis, Schüler, et al